In December 1947, the first generation of human semiconductor amplifiers was born in Bell Laboratories.
It was this small transistor that continued to rewrite the world in the following 75 years. At the same time, the development of the transistor itself also entered a bottleneck, and Moore’s Law slowed down.
In the 75th year since the birth of the transistor, what methods can be used to continue Moore’s Law ?
In 2022, we still need new transistors ?
Does our world still need better transistors?
With the improvement of transistors, humans have mastered new capabilities that have never been imagined, such as computing and high-speed communications, the Internet, smartphones, memory and storage, computer technology, and artificial intelligence, and it is conceivable that there will be more in the future Other new technologies emerge;
The wide application of transistors is changing all technologies, industries, and sciences. At the same time, the evolution of semiconductor technology is not limited by its material and energy usage like other technologies. ICs can be produced using relatively few materials and are becoming more and more are getting smaller and using less material, and the ICs themselves are getting faster and more efficient;
In theory, the energy required for information processing can still be reduced to less than one-thousandth of the energy required today. Although we may not yet know how to achieve this theoretical efficiency, we know that it is theoretically possible. The energy efficiency of most other technologies has already reached the theoretical limit.
In 2030, a single chip can hold 1 trillion transistors
It is a fact that new transistors are needed, but the development and manufacture of new transistors have been difficult, both economically and technically, and new difficulties have been encountered.
The development path of transistor technology is inherently uneven, and there are huge challenges to be dealt with almost every once in a while.
Around 1980, the dynamic power consumption of chips became a big problem. After replacing NMOS and bipolar technology with CMOS, the working voltage was reduced from 5 volts to 1 volt, which brought great progress;
From 2000 to 2010, the static power consumption of the chip became a challenge again. According to the prediction of the researchers at that time, the heat generated by IC per square centimeter would soon reach the heat of the nuclear reactor core, but later 3D Fin-FET and multi-core processors The architecture solved this problem, and the development of transistors has entered a relatively stable development period.
Up to now, the performance improvement and power consumption reduction brought about by the progress of Fin-FET are becoming more and more limited. The industry is using a new 3D CMOS structure gate-all-around (GAA) to manufacture new transistors, and Intel is one of them. member.
Not long ago, to further reduce the three-dimensional size of transistors, Intel realized GAA with the structure of RibbonFET but found that when the distance between the source and drain is further reduced, it will produce a more obvious short channel effect and leakage.
This situation can be improved if the traditional channel material silicon is replaced with a new material other than silicon. There are also some related studies in academic circles, using a material called transition metal sulfide as a channel material.
This material is only three atoms thick and has good electron mobility. It has natural advantages as a channel material.
In terms of this 2D material, Intel has also done a lot of research and analysis on this material, and at the meeting demonstrated a full-surround gate stacked nanosheet structure, using a 2D channel material with a thickness of only three atoms, while achieving near-ideal low-leakage double-gate structure transistor switching at room temperature.
In addition, 3D packaging technology can further increase the number of transistors in a single device.
Intel has also made new progress in 3D packaging.
Compared with the results announced at IEDM 2021, the latest hybrid bonding research presented at Intel IEDM 2022 has increased power density and performance by another 10 times.
In addition, by continuing to shrink the interconnect pitch to 3 microns through hybrid bonding technology, Intel has achieved interconnect density and bandwidth similar to monolithic system-on-chip (system-on-chip) connections.
In addition, the materials required in the multi-chip interconnection process are replaced with inorganic materials, to be compatible with various process requirements of packaging factories.
Although it is a matter of huge financial and manpower to further realize the miniaturization of transistors, there are still companies like Intel that continue to invest in research and development and have expectations for the future of transistors.
Intel believes that from 2023 to 2030, the number of transistors in a single device will increase tenfold, from 100 billion transistors to 1 trillion transistors.
To achieve this goal, industry-leading companies such as Intel need to continue to invest in research and development and try more feasible technologies.
Frequently Asked Questions (FAQ)
Q: How are millions of transistors put on a chip?
A: The process of putting millions of transistors on a chip is called photolithography. It involves using a photo mask and light-sensitive chemicals to transfer a circuit pattern onto a silicon wafer. The wafer is then etched to remove unwanted material and create the tiny transistors. This process is repeated multiple times to create layers of circuitry on the chip, with each layer connected by tiny wires called interconnects. The end result is a small, highly complex piece of circuitry that can perform a wide range of tasks.
Q: What is a transistor and why are there millions of them on a chip?
A: A transistor is a tiny electronic component that acts as a switch or amplifier in a circuit. It is made up of a semiconductor material, typically silicon, and is used to control the flow of electricity in a circuit. Millions of transistors are put on a chip because they allow for much more complex and powerful circuitry to be created in a small space. This enables computers, smartphones, and other electronic devices to be much faster and more powerful than they would be otherwise.
Q: How does photolithography work in the process of putting transistors on a chip?
A: In photolithography, a photo mask is used to project a pattern onto a silicon wafer that has been coated with a light-sensitive chemical called a photoresist. The photoresist is then exposed to light, which causes it to harden in the areas that were exposed to the light. The wafer is then etched to remove the unwanted photoresist, leaving behind the circuit pattern. This pattern is then used as a guide to etch the underlying silicon, creating the tiny transistors.
Q: Can the number of transistors on a chip continue to increase indefinitely?
A: In the past, the number of transistors on a chip has increased exponentially, following what is known as Moore’s Law. However, there are physical limitations to how small transistors can be made, and it is becoming increasingly difficult and expensive to continue to shrink them. As a result, the rate at which the number of transistors on a chip is increasing is slowing down. However, there are still new technologies being developed that may allow for further increases in the number of transistors on a chip.